SM, I have used several optimized results for a yagi antenna, but I have never worked with a quad. I found that at the extremes of optimized modeling the results may look good on screen or paper, but when applied in a real world setting I often ended up with beam real performance that was almost omni directional instead of directional. Of course I could only judge these results via my human senses and the results that I saw on my receive meter. I also did some on-air comparisons with some friends as well, but those results were also not very impressive.
I have no idea what real results you are seeing when you apply the optimized results to the construction of your beam, but:
1. I would suggest that your using only 11 segments is not serving your design and optimization needs very well.
2. I also discovered that pressing the limits of the software by re-optimizing an already optimized model, if that is what you're trying to doing, was probably exceeding the technical limits for accuracy in the software. The side lobe results you describe give me that impression at least.
Good luck and keep us posted.